Efficient time-dependent fatigue reliability assessment accounting for material variability in steel structures

For steel structures with highly variable number of cycles to failure and when the load history is the output of a complex stochastic process, assessing fatigue damage with time is challenging. In this article, non-linear accumulation damage model available in literature is considered. With this model, the history of loading, that is to say the order or cycles, is taken into account. Using this model to estimate probability of failure requires great computational costs. In this article, two numerical strategies to reduce these costs are proposed: by creating kriging-based meta-models and defining relevant enrichment procedures, an estimation of the instantaneous probability of failure is computed. In addition, upper and lower bounds of the probability of failure are derived and enable to assess the quality of the simulation. Numerical illustrations are given on a plate with material variability and 10 000 load cycles.